Discussion
Pollen morphology of species of Sorbus s.lat., Aria and their hybrids in Norway (Hedlundia ), is rarely studied in detail. The first European studies of pollen morphology of Sorbus in general, was Erdtman (1952) and Praglowski (1963), later followed by Eide (1981). These studies were novel as they believed thatSorbus pollen could be identified, and not only categorized asCrataegus -type of pollen as earlier. Later, Boyd and Dickson (1987) studied pollen of S. aucuparia (2n), A. rupicola(4n), H. arranensis (3n), and H. pseudofennica (4n), using LM. They found that S. aucuparia could be separated from the other taxa but found less differences between the hybrid species. Bednorz et al. (2003) found that while size measures were overlapping for species and hybrids of Sorbus , exine sculpturing provided features for taxonomic classifications. There has been a general opinion amongst palynologists that pollen from the family Rosaceae only is useful for morphological identification on higher ranks, like subfamily or genus. Wronska-Pilarek et al. (2022) points out that studies report useful morphological characters for species delimitation in the Rosaceae, such as the length of the polar axis (P), pollen shape (P/E), operculum structure, and presence/lack of costae colpi.
Bednorz et al. (2005) concludes that “… it is relatively easy to separate pollen grains” of the genera Sorbus ,Chamaemespilus , and Torminalis using SEM. They found that type of sculpture, length of polar axis (P), number of ectocolpi, and ratio of the ectexine to endexine thickness are the most important separating characters for these genera. These taxa were previously defined within the genus Sorbus but are now classified as genera (Sennikov and Kurtto, 2017).
Our results show that pollen of Aria generally has a longer polar axis (P) than pollen of Sorbus and Hedlundia (Fig. 3). From the PCA analysis (Fig. 7), Aria has clearly larger pollen size in general, compared to Sorbus and Hedlundia . AsAria and Sorbus are not closely related (Lo and Donoghue, 2012), but Aria is related to Malus , whereas Sorbusare related to Pyrus , it is expected to see differences between these genera, and the main separating character is size, and more specific, the length of the longest axis (P). For the hybrids (Hedlundia ), it has been hypothesized that Sorbus is the ovule donor (Liljefors, 1955), and therefore it’s reasonable to believe that pollen mostly resembles their mother in size.
Pollen size was categorized in different intervals by Erdtman (1969). Pollen with polar axis (P) between 50-100 µm are large, 25-50 µm are medium sized, and between 10-25 µm are defined as small. From our measurements, A. edulis , H. hybrida 1 and A. obtusifolia have a mean P value above 25 µm and are therefore medium sized. All other specimens in our study have small pollen. With its medium sized pollen, H. hybrida 1, stands out as very large compared to all other species of Hedlundia . H. hybrida is a morphologically very variable species, and different subspecies have been suggested (Grundt and Salvesen, 2011). This variation is also seen here, when comparing the two different specimens of H. hybrida in our study. In the other end of the range are H. neglecta , with very tiny pollen.
We have compared our measurements of the shape of pollen (ratio P/E) with other available studies and see that there is a large range between studies (see Figure 8). In general, our samples have a higher ratio than in all the other studies, whereas Eide (1981) has lower ratios than the others. This can be related to different preparation techniques. According to Katsiotis and Forsberg (1995) pollen size can be influenced by treatment methods and mounting media. Pollen grains tend to swell in glycerol (Moore and Webb, 1978). This can lead to different sizes in different studies and can explain why our measured pollen has a higher ratio as the samples are all mounted in glycerol.
S. aucuparia pollen ranges from a ratio of 0.96 in Eide (1981) to 1.48 in this study (Figure 8). This gives rise to all forms of pollen; oblate spheroidal form (Eide, 1981), prolate-spherioidal shape (Bednorz et al., 2003; Boyd and Dickson, 1987; Erdtman et al., 1961), subprolate shape (Bednorz et al., 2005; Praglowski, 1963), and prolate shape in this study. S. aucuparia is a morphological extreme variable species, and many have suggested establishing subspecies to account for this variation. This may also be reflected through this study of pollen shape.
For H. hybrida , the variation in P/E is lower, ranging from 1.05 to 1.36, but still giving rise to three different shape forms. H. hybrida is also a very variable species, resulting from hybridization events with different parent trees. Our two samples of H. hybridaoriginates from different parts of Norway, H. hybrida 1 from Telemark (4), eastern Norway, whereas H. hybrida 2 is collected in Sogndal (5) in western Norway and is a morphological very different type. From the comparison of different data on the shape ofSorbus pollen, we can conclude that shape is only useful where the methodology is consistent.
Erdtman (1952) hypothesized that pollen size is dependent on polyploidy. Erdtman (1969) writes that “Pollen grains and spores in polyploid species are, as a rule, larger than in diploids”. In general, it is argued that larger amount of DNA per cell will make larger cells, both for pollen and spores. Several examples of species with larger pollen size correlated with ploidy levels are shown, as for Andropogonby Gould (1957), for Avena species studied by Katsiotis and Forsberg (1995), and Altmann et al. (1994) even used pollen size to determine the ploidy level of Arabidopsis thaliana .
After the “Eukaryotic genome size databases” (Gregory et al., 2007) was launched, Knight et al. (2010) used this database together with pollen size data from different sources to study the relation between pollen and genome size. They found that from 464 species, there was a positive correlation between pollen width (E) and genome size. Based on literature, Knight et al. (2010) estimated that pollen size increase by 1.1x-2x with double C-value (amount of DNA in a haploid genome).
When plotting our mean pollen width (E) against ploidy level (DNA amount) we find a positive correlation of 0.45, as shown in Figure 2 (small box). This supports the hypothesis that more DNA results in larger pollen, hence pollen of polyploids are thicker than diploids.
Eide (1981) argued that exine ornamentation was one of the most important features not only to identify the pollen of Sorbus , but also the Rosaceae. Bednorz et al. (2003) finds that the exine sculpturing was the main character for distinguishing pollen fromS. aucuparia , H. arranensis , H. hybrida , andH. mougeotii , and concluded that S. arranensis was distinct from the other species by its short rugulate-like exine pattern. Both S. hybrida and S. mougeotii were characterized by very long striation, with the striae clearly thinner atS. mougeotii . The S. aucuparia pollen had shorter striae and dense, well visible perforations between vallae.
The variation in ornamentation of our samples is shown in the SEM pictures given in Figure 5. The variation in surface structures is clearly visible and a basis for classification of the taxa included in this study. Perforations is variable, and a basis of our key to identify Norwegian species:
Medium sized pollen (25-50 µm)
Few perforations Aria obtusifolia
Medium perforations Aria edulis
Dense perforations Aria rupicola
Small pollen (10-25 µm)
Few perforations
Polar view circular Medium length of striae Hedlundia sognensisShort striae Hedlundia subpinnata Polar view triangular Hedlundia neglecta
Medium perforations
Short striae Hedlundia hybridaLong striae Hedlundia mougeotii
Dense perforations
Circular polar view Short striae in parallel directions Hedlundia subarranensisMedium striae in variable directions Hedlundia subsimilisLong striae in parallel directions Sorbus aucuparia Triangular polar view Striate-rugulate exine Hedlundia lancifoliaStriate exine Hedlundia meinichii
It is possible to identify pollen of Sorbus , Aria andHedlundia with morphological characters using LM only, but using SEM will ease the identification. The identification will need a very close study of the given characters but will be feasible.
This study of pollen morphology of Sorbus and Aria with their hybrids (Hedlundia ), shows that the genus Aria has longer pollen than both Sorbus and Hedlundia , and that these are separated by size. Sorbus pollen is smaller thanHedlundia but with some overlap. By using size in combination with several characters of surface ornamentation, hybrids are possible to identify by pollen morphology.